AI Policy
Following ISO/IEC 42001:2023 · Last updated:
Our standards status. Easy HR follows the practices of ISO/IEC 42001:2023, Information technology — Artificial intelligence — Management system, the international standard for managing AI responsibly. Easy HR is not certified to this standard: no certification body has audited us against it. We use it as the framework for how we develop, use and oversee AI, as this policy describes.
We welcome feedback on our services. It helps us improve our ISO standards research and the way we apply the standard. Email support@aifoundation.net.au.
1. Purpose and scope
This policy sets out how AI Foundation, which runs Easy HR, develops, uses and oversees artificial intelligence (AI). It is our AI policy under ISO/IEC 42001:2023, and it applies to:
- every AI feature in Easy HR, now and in future;
- everyone who builds, runs or administers Easy HR, including our SuperAdmins; and
- the suppliers whose AI services we use.
It sits alongside our Information Security Policy, Privacy Policy and Terms of Service.
2. Our commitments
We commit to:
- help people, not replace their judgement: people make the decisions that matter;
- treat everyone fairly, and never let AI choose who gets a job by age, sex, race, religion, disability or any other ground the law protects;
- protect privacy, giving AI only the information each task needs;
- be open about where and how we use AI;
- make our AI reliable and secure, and check that it works as intended;
- meet our legal obligations, including Australian privacy, anti-discrimination and consumer law;
- manage AI risks and impacts through each feature's whole life, from design to retirement; and
- keep improving, by learning from feedback, incidents and reviews.
We also follow Australia's AI Ethics Principles.
3. Where Easy HR uses AI
Easy HR has four AI features. Each one uses Google's Gemini AI service, called from our servers and never from your browser.
- Resume reader. Reads a resume that a jobseeker adds, and builds their profile: work categories, industries, fields of work, job titles and descriptions, duties, skills, qualifications, experience and locations.
- Search reader. Reads what a company or HR Agent types in the Search Candidates box, from a few words to a whole job ad, and sorts it into requirements.
- Meaning search. Turns the anonymous summary of each jobseeker's work, and each search, into a fingerprint of its meaning (an embedding), so search can find people who describe the same work in other words.
- Candidate matcher. Compares those requirements with jobseekers' work profiles, and gives each candidate a match score with reasons. The score is then adjusted for distance.
If the AI service is unavailable, search still works, using our own rule-based scoring.
We don't use AI to approve or decline registrations, to decide who is hired, or to make any other decision that significantly affects someone without a person involved.
4. Human oversight
- Jobseekers review their AI-built profile and can change anything before they submit it, and edit it later.
- SuperAdmins check resumes and profiles before approving anyone as a JobSeeker, and can correct a profile at any time.
- Match scores are suggestions. Companies and HR Agents see the reasons for each one, and make their own decisions.
- Anyone can ask for a person to review how our AI has described them, by contacting us.
5. Fairness and lawful recruitment
- Our search sets aside requirements based on age, sex or gender, pregnancy, relationship status, sexual orientation, race, colour, ethnicity, nationality, religion, political opinion, disability, health, family or carer's responsibilities, or union membership, and shows the searcher what it set aside. Genuine requirements of a job, such as a licence or the right to work in Australia, are allowed.
- Our matcher judges only work: skills, experience, qualifications, licences and duties.
- Our resume reader is instructed never to guess anyone's age, gender, ethnicity, religion, health or disability, and to keep private details out of the summary companies see.
- We check AI results for signs of unfair treatment, and fix what we find.
6. Data and privacy
- AI receives only what each task needs. When comparing candidates, it sees an anonymous summary of each person's work: never a name, contact details, address, employer, referee, ID number or demographic details.
- Meaning search fingerprints are made from that same anonymous summary, never from where someone lives. They're kept on our servers with each profile, can't be read from the app, are made again when a profile changes, and are deleted when it is.
- We don't train AI models with your information.
- AI answers must follow a fixed structure, and we clean them before we use them: anything outside that structure is dropped.
- Personal information is handled under our Privacy Policy.
7. Risks and impacts
Before we release an AI feature, and whenever we change one significantly, we assess:
- the risks, such as inaccurate profiles, unfair results, privacy breaches, misuse, over-reliance on match scores and failures of the AI service;
- the impacts on the people affected (jobseekers, companies, HR Agents and the wider community), including on people's access to work; and
- the controls that reduce those risks to an acceptable level, such as the human checks, data limits and fairness rules in this policy.
We record our assessments, and review them when something changes.
8. Building, testing and monitoring
- Each AI feature has a defined purpose, inputs and outputs, and a list of things it must never do.
- We test AI features with automated tests before every release, including tests that keep private details out of what the AI sees and what companies see.
- AI features have usage limits for each person, to prevent misuse.
- We monitor how our AI features perform, including errors and feedback, and fix problems promptly.
- Our AI instructions and code are kept in version control, so every change is tracked.
- When we retire an AI feature, we remove its access to information and keep the records we need.
9. Transparency
- We tell you where Easy HR uses AI, in this policy, in our Privacy Policy and in the app where AI helps, such as AI-built profiles and match scores.
- Match scores come with reasons, and searches show what was set aside.
- If an AI incident affects you, we'll tell you.
10. AI suppliers
- We use Google's Gemini AI service. We choose AI suppliers with strong security and privacy commitments, and review their terms when they change.
- We remain responsible for how Easy HR uses AI, whichever service powers it.
11. Roles and responsibilities
- AI Foundation's leadership approves this policy, provides the resources to follow it, and is accountable for Easy HR's AI.
- SuperAdmins review resumes and AI-built profiles, approve registrations, and act on concerns.
- Everyone who works on Easy HR follows this policy and reports concerns.
- Users check AI results and use AI features lawfully, as our Terms of Service require.
12. Concerns, incidents and feedback
Tell us at support@aifoundation.net.au if you think our AI has got something wrong, treated someone unfairly or been misused. We'll acknowledge your report, investigate it, fix the cause, and tell you what we did.
We're not certified to ISO/IEC 42001, and we're always learning. We welcome feedback on our services: it helps us improve our ISO standards research and the way we apply the standard.
13. How we follow ISO/IEC 42001
ISO/IEC 42001 sets out how an organisation should run an AI management system. This is how Easy HR follows its main parts:
- Context of the organisation (clause 4): who our AI affects and what they need, and the features this policy covers (section 3).
- Leadership (clause 5): this AI policy, and the roles in section 11.
- Planning (clause 6): risk and impact assessments, and our objectives for fair, accurate and private AI (section 7).
- Support (clause 7): the people, skills, tools and records our AI relies on, and awareness of this policy.
- Operation (clause 8): how we build, test, release and run AI features (section 8).
- Performance evaluation (clause 9): monitoring, feedback and regular review (sections 8, 12 and 14).
- Improvement (clause 10): fixing problems at their cause, and improving continually.
Annex A of the standard lists controls in nine areas. We apply them as follows:
- Policies related to AI: this policy.
- Internal organisation: roles, and how concerns are reported (sections 11 and 12).
- Resources for AI systems: each feature's data, tools, systems and people (section 3).
- Assessing impacts of AI systems: section 7.
- AI system life cycle: section 8.
- Data for AI systems: section 6.
- Information for interested parties: section 9.
- Use of AI systems: sections 3 to 5.
- Third-party and customer relationships: section 10, and our Terms of Service.
14. Reviewing this policy
We review this policy at least once a year, and whenever we add or significantly change an AI feature, after a significant incident, or when the law or the standard changes. The date at the top shows the latest version.
15. Contact us
AI Foundation (Easy HR)
Email: support@aifoundation.net.au
Website: easyhragency.com